Executes a single parameter combination from the parameter grid. This function is called by PBS array job workers.
Usage
run_single_task(
task_id,
param_grid_file,
output_dir,
sweep_type = c("modelsel", "baselines"),
methods = c("MFA", "MPCA"),
strategies = c("grid", "twostage"),
K_candidates = 1:5,
r_candidates = 1:5,
mode = c("modelsel", "oracle_r", "oracle_Kr"),
criterion = c("ICL", "BIC"),
icl_temperature = 2,
max_iter = 50,
tol = 0.001,
n_init = 1,
use_kmeans_init = TRUE,
align_basis = TRUE,
refine_assignments = TRUE,
verbose = TRUE
)Arguments
- task_id
Integer. The task ID corresponding to run_id in the parameter grid.
- param_grid_file
Character. Path to the RDS file containing the parameter grid.
- output_dir
Character. Directory to save results.
- sweep_type
Character. Type of sweep: "modelsel" or "baselines".
- methods
Character vector. Methods to evaluate (e.g., c("MFA", "MPCA")).
- strategies
Character vector. Selection strategies (e.g., c("grid", "twostage")).
- K_candidates
Integer vector. Candidate K values for model selection.
- r_candidates
Integer vector. Candidate r values for model selection.
- mode
Character. Selection mode: "modelsel", "oracle_r", or "oracle_Kr".
- criterion
Character. Model selection criterion: "ICL" or "BIC".
- icl_temperature
Numeric. Temperature parameter for ICL entropy scaling.
- max_iter
Integer. Maximum EM iterations.
- tol
Numeric. Convergence tolerance.
- n_init
Integer. Number of random initializations.
- use_kmeans_init
Logical. Whether to use k-means initialization.
- align_basis
Logical. For baselines sweep: whether to align basis vectors.
- refine_assignments
Logical. For baselines sweep: whether to refine cluster assignments.
- verbose
Logical. Whether to print progress.